Frequent changes in subtelomeric DNA methylation patterns and its relevance to telomere regulation during human hepatocarcinogenesis

Subtelomeric chromatin modifications are important regulators of telomere length. We examined the subtelomeric DNA methylation status of 7q, 8q, 17q, 18p, 21q and XpYp in 32 pairs of hepatocellular carcinomas (HCCs) and their adjacent non‐HCCs via methylation‐specific PCR (quantified as methylation ratio). In addition, 10q was subjected to bisulfite‐genomic‐sequencing. Telomere length was determined by Southern hybridization. In all cases, the relationship between methylation ratio and telomere length was determined. High levels of methylation ratio were found on chromosomes 7q, 18p and XpYp, whereas 8q 17q and 21q were less methylated in both HCCs and non‐HCCs. Compared to non‐HCCs, HCCs exhibited a higher methylation ratio on 18p and 21q, and a wider distribution of methylation ratio on 7q, 21q and 10q (p < 0.05). The methylation ratio of 18p and of 21q was negatively and positively correlated with telomere length of HCCs, respectively (p < 0.05). We evaluated changes in methylation pattern between non‐HCCs and HCCs. Out of 185 sites, hypermethylation changes from non‐HCC to HCC were found at 47 sites and hypomethylation changes at 31 sites. Changes in methylation pattern were observed at three to four sites among six chromosomal sites in 15 patients (47%). There was a tendency toward hypomethylation changes at 7q (p = 0.013) and hypermethylation changes at 21q (p = 0.057) when telomere lengthened from non‐HCCs to HCCs. In summary, subtelomeric methylation patterns dynamically changed during hepatocarcinogenesis. Subtelomeric methylation at certain regions was related to telomere lengthening or shortening, suggesting an association between subtelomeric chromatin structure and telomere length regulation in human hepatocarcinogenesis.

A telomere, which is a specialized structure composed of tandem repeats of DNA and associated proteins, is crucial for chromosome integrity. 1 Dysfunctional telomeres result in chromosome end-to-end fusion, which leads to chromosomal instability. [2][3][4][5] Telomere length is maintained by at least two mechanisms: telomerase and homologous recombination. Telomerase adds telomeric repeat sequences to the 3 0 end of telomeres. 6 Most cancer cells possess active telomerase, and thus, telomere length is maintained during successive proliferation. 7,8 In telomerase-negative cancer cells, telomere is maintained by telomere recombination, which results in long telomeres with heterogeneity. This mechanism is known as alternative lengthening of telomeres (ALT). 9 Mapping and sequencing have been performed in detail on human subtelomeric sequence assemblies within a distance of 500 kb of the telomere. 10 A total of 941 transcripts, including 214 single-copy genes, were found in subtelomeric sequence assemblies, and the transcript density is similar to that found genome wide. 11 Similar to pericentromeric regions, subtelomeres have a high density of DNA repeats and CpG islands, which are susceptible to DNA methylation by DNA methyltransferases (DNMTs). 12,13 Telomeres, however, do not contain CpG sequences.
Recent evidence indicates that epigenetic modification of subtelomeric and telomeric chromatin exerts influence on the regulation of telomere length. 14 In DNMT-deficient cells, demethylation of subtelomeric regions induced telomere elongation, which was suggested to result from increased homologous recombination in telomeric sequences. 15 In telomerase knockout mouse cells, telomere shortening was associated with features of open chromatin, revealing decreased DNA methylation and increased histone acetylation in subtelomeres. 16 These data suggest that subtelomeric DNA methylation is implicated in telomere metabolism.
Hepatocellular carcinoma (HCC) is the seventh most common cancer worldwide, and more than 80% of HCC cases are caused by chronic hepatitis B virus (HBV) infection. 17 Telomere length becomes shorter in the early stages of human hepatocarcinogenesis, 18,19 and these short telomeres are thought to be one of the factors that can induce chromosomal instability, an important mechanism of HBV-related hepatocarcinogenesis. 20,21 Epigenetic modification has been implicated as a critical event in carcinogenesis. 22 Aberrant DNA methylation of proto-oncogenes and tumor suppressor genes has frequently been found in human cancers. [23][24][25] Epigenetic modifications of human subtelomeres, however, have not been characterized in tumors. Moreover, their impact on telomere length regulation and telomere function has not been examined during tumorigenesis, particularly during hepatocarcinogenesis.
In this study, we examined the subtelomeric methylation status at six CpG islands in 32 pairs of HCCs and their adjacent noncancerous liver tissues (non-HCCs) using methylationspecific PCR (MSP), and at one CpG island in 19 pairs of HCCs and their adjacent non-HCCs using bisulfite genomic sequencing (BGS). The subtelomeric methylation pattern was dynamically altered during hepatocarcinogenesis, and the methylation ratios of 18p and 21q correlated with telomere length in HCCs, and change in methylation of 7q and 21q correlated with telomere length changes during hepatocarcinogenesis.

Material and Methods
Tissues Thirty-two HCCs and their adjacent non-HCCs were sampled from resected liver specimens, frozen directly in liquid nitrogen, and stored at À80 C. The study included 26 male and six female patients with an age range of 40 to 67 years (mean 6 SD ¼ 53 6 7.98). Twenty-eight patients were positive for HBV, one had hepatitis C virus (HCV), one patient was an alcoholic and two had none of these conditions. Liver resection was performed due to metastatic colon cancers. The noncancerous liver tissues showed normal histology, except for mild steatosis. The specimens were supplied by the Liver Cancer Specimen Bank of the National Research Resource Bank Program of the Korea Science and Engineering Foundation (Ministry of Science & Technology, Republic of Korea).

Genomic DNA extraction
A 20-to 40-mg sample of frozen tissue was ground using a pestle in liquid nitrogen and treated with 700 lL of lysis buffer containing 20 lg/mL protease K (Cosmo Genetech, Seoul, Korea), 20 mM TrisÁHCl (pH 8.0), 5 mM EDTA (pH 8.0), 400 mM NaCl and 1% SDS. Samples were incubated overnight at 42 C in a shaking water bath and then extracted three times with phenol or chloroform. After extraction, genomic DNA was isolated by isopropanol precipitation. The DNA pellet was washed with 70% ethanol, air-dried and then dissolved in 300 lL of water. DNA integrity and RNA contamination were checked by gel electrophoresis.

Bisulfite modification of genomic DNA
Bisulfite modification of genomic DNA was performed using the EZ DNA methylation kit (Zymo Research Corp., Orange, CA) following the manufacturer's instructions. Specifically, 1 lg of genomic DNA was mixed with 5 lL of M-dilution buffer in a total volume of 50 lL, and the sample was incubated at 37 C for 15 min. Cytosine-to-thymidine conversion was followed by adding 100 lL of CT conversion reagent, which was then incubated at 50 C for 15 hr. Modified DNA was purified via Zymo-Spin 1 Columns, dissolved in 30 lL of purified water and used immediately or stored at À20 C until use.

Methylation-specific PCR
MSP primers were designed with MethPrimer (http:// www.urogene.org/methprimer/). The primer sequences for both methylated and unmethylated regions are provided in Table 1. CpGenome TM Universal Methylated DNA (Chemicon, Temecula, CA) was used as a control. The PCR mixture contained 1 lL of bisulfite-modified genomic DNA, 1.5 lL of 10 lM forward primer, 1.5 lL of 10 lM reverse primer, 2.5 lL of 10Â PCR buffer, 2 lL of 25 mM MgCl 2 , 2 lL of 2.5 mM dNTP and 0.2 lL of AmpliTaq Gold (Applied Biosystems, Foster City, CA) in a total volume of 25 lL. PCR reactions began at 95 C for 10 min and were followed by 40 cycles of amplification [denaturation at 95 C for 30 sec, annealing for 30 sec (temperatures are given in Table 1), and extension at 72 C for 30 sec]. A final extension at 72 C for 5 min concluded the reactions. PCR products were separated by electrophoresis on 2% agarose gels.

Bisulfite sequencing analysis
Primers for BGS were designed using MethPrimer. The primer sequences were 5 0 -GGA GTT TAG GAT TTA GAT TTG GTT TTA G-3 0 (sense) and 5 0 -AAC AAA TTA AAA AAT CCC CTC TAC C-3 0 (antisense), resulting in a 298-bp product from chromosome 10q. These primers were specific for modified DNA but did not contain any CpG sites in their sequence, enabling both methylated and unmethylated DNA to be amplified by the same primer set. PCR was performed under the same conditions as MSP with the following changes: 5 lL of bisulfite-modified genomic DNA, 1 lL of forward primer, 1 lL of reverse primer, denaturation at 94 C and annealing at 55 C. The PCR product was subjected to Tvector cloning (Invitrogen, Carlsbad, CA). Cloned DNA was column-purified (Qiagen, Hilden, Germany), and two to nine clones were randomly chosen for automated sequencing (Cosmo Genetech, Seoul, Korea).

Telomere terminal restriction fragment length analysis
Telomere length was measured using Southern hybridization, 19,26 and the mean telomere length was calculated as described by Kruk et al. 27 Briefly, 10 lg of genomic DNA was digested with Hinf I overnight, purified by phenol extraction and isolated by ethanol precipitation. DNA concentration was remeasured using spectrometry, and 2 lg of Hinf I-digested DNA was fractionated on a 0.7% agarose gel. The gel was denatured in 1.5 M NaCl and 0.5 M NaOH for 30 min twice, and then neutralized in 1.5 M NaCl and 0.5 M Tris-HCl, pH 7.5 for 30 min. Then, DNA was transferred to a nylon membrane using upward capillary transfer in 10Â SSC (1.5 M NaCl/0.15 M sodium citrate) overnight. Hybridization was carried out with a 3 0 -end DIG-labeled d(TTAGGG) 4 probe (Roche Molecular Biochemicals, Mannheim, Germany) at 37 C for 12 hr, and DNA detection was performed according to the manufacturer's instructions. The resulting X-ray film was scanned with a luminescent image analyzer (Fujifilm, Tokyo, Japan), and the telomere signal in each lane was quantified in a grid object, which was defined as a single column with 25 rows (Image Gauge Software 2.54, Fujifilm). 27 The lane corresponding to size markers was also applied to the grid object, and the row number against the log 10 (molecular size) was plotted to determine the molecular size for each of the 25 rows. 27 Telomere length was calculated as follows: where ODi is the densitometer output and MWi is the length of the DNA at position i. 27 Sums were calculated over the range of 2.0-23 kb.

Statistical analysis
All statistical analyses were performed using SPSS 13 (SPSS, Chicago, IL). Data were analyzed with two-sided p values from a 2 Â 2 contingency table determined by a v 2 test and by Levene's test for equality of variances. Significance was set at a p value <0.05.

Primer design for MSP and BGS
For the design of MSP and BGS primers, we selected CpG islands at chromosomes that are known to have no sequence gaps and no telomere-like repeat sequences at subtelomeric regions. 10,11 MSP was performed on chromosomes 7q, 8q, 17q, 18p and 21q and XpYp at locations 104, 27, 4, 2, 31 and 40 kb from telomeres, respectively. Figure 1 shows a representative MSP result for chromosome 7q with a map of MSP primers. BGS, which provides a quantitative methylation ratio, was performed on chromosome 10q at a location 26 kb upstream of the telomere, and a map of BGS primers and results are shown in Figure 2.

Subtelomeric methylation in HCCs and non-HCCs
MSP was determined at six chromosomal regions in 32 pairs of HCC tissues (192 sites total) and adjacent non-HCC samples (192 sites total). Methylation was quantified by measuring the intensity of the DNA bands. The methylation ratio (%) of each sample is summarized in Table 2.
Next, we compared the methylation ratio between HCCs and non-HCCs in individual regions (Fig. 3a). Methylation
BGS performed on chromosome 10q in 19 pairs of HCCs and their adjacent non-HCCs showed that methylation frequency ranged from 44 to 76% in non-HCCs and from 7 to 93% in HCCs (Fig. 2b). HCCs showed greater variance in methylation frequency (p ¼ 0.003; Fig. 2c) although the mean value of methylation frequency (58 6 21.9) was similar

Cancer Genetics
Oh et al.
to that found in non-HCCs (62 6 8.2). Ten HCCs were found to have higher methylation ratio than their adjacent non-HCCs, and nine HCCs had lower ratios than the adjacent non-HCCs (Fig. 2b).
We evaluated differences in methylation ratio of individual chromosomes between HCCs and their adjacent non-HCCs (Fig. 3b). The difference in methylation ratio was determined as follows: methylation ratio of HCC À methylation ratio of adjacent non-HCC. The difference in methyla-tion ratio of 17q and 21q widely dispersed, and 8q, 18p and XpYp showed a narrow range of the difference (Fig. 3b).
The sites with methylation change were evaluated in each patient. Changes in methylation pattern occurred at more than two out of six sites in most HCC patients (29/32, 91%). Eleven patients were found with changes in methylation at three sites, and four patients at four sites, and one patient showed no changes at all tested sites (Figs. 4a and 4d).

Subtelomeric DNA methylation ratio and telomere length in HCCs and their adjacent non-HCCs
The telomere length of 32 pairs of HCCs and non-HCCs was measured by Southern blot analysis (Fig. 5a), and the results  were summarized in Figure 4a. The telomere length in HCCs ranged from 4.5 to 12.7 kb, with the mean of 8.2 6 2.39 kb, and non-HCCs had a telomere length from 5.6 to 11.7 kb, with the mean of 8.7 6 1.20 kb (Fig. 4a).
We analyzed the correlation between the methylation ratio and telomere length in HCCs and non-HCCs (Fig. 5b). The methylation ratio of 18p negatively correlated with telomere length in both HCCs (p ¼ 0.037) and non-HCCs (p ¼ 0.084), although significance was marginal in non-HCCs (Fig.  5b). A similar result was seen in 7q in HCCs with marginal significance (p ¼ 0.077) (Fig. 5b). Meanwhile, chromosome 21q in HCCs showed a positive correlation with telomere length (p ¼ 0.025). Chromosomes 8q, 17q, YpYp and 10q, however, showed no significant correlation.
Next, we examined whether the difference in methylation ratio between HCCs and non-HCCs was related to telomere length changes during hepatocarcinogenesis (Fig. 6). Methylation changes negatively correlated with telomere length changes in 7q (p ¼ 0.013); methylation changes proceeded further toward hypomethylation as telomere lengthened from non-HCCs to HCCs. Conversely, there was a tendency toward hypermethylation changes in 21q as telomere length

Discussion
Subtelomeres and telomeres are generally heterochromatic. 15,28,29 This study showed that subtelomeric methylation status appeared to vary from region to region on six chromosomes in HCCs and their adjacent non-HCCs. Most samples exhibited high levels of methylation in 7q, 18p and XpYp, and relatively low levels in 8q, 17q and 21q. This suggests that chromatin structure varies among subtelomeric domains, and that both compacted and relaxed chromatin structures coexist at subtelomeres.
Overall subtelomeric methylation status was similar in HCCs and their adjacent non-HCCs; however, at individual regions, HCCs in 18p and 21q had higher methylation ratio, and HCCs in 7q, 21q and 10q showed a wider distribution of methylation ratio compared to non-HCCs. Moreover, methylation patterns appeared to have been dynamically altered from non-HCCs to HCCs. In fact, methylation patterns were different between HCCs and their adjacent non-HCCs at a frequency of 42% (78/185 sites). Fifteen patients (15/32, 47%) appeared to have methylation pattern changes at three to four sites among the six chromosomal sites. These results indicated frequent methylation changes in subtelomeric regions during hepatocarcinogenesis. The MSP and BGS sites examined in this study are located far from the genes; thus, alteration of subtelomeric DNA methylation is assumed to exert little effect on the expression of adjacent genes. Instead, frequent methylation changes in subtelomeric regions might affect local chromatin structures. Telomeric repeat-containing RNA (TERRA) is known to have a function in orchestrating chromatin remodeling, 30 and TERRA transcription is regulated by CpG methylation at their promoters, 31 suggesting that there might be putative connection between subtelomeric Figure 6. Relationship between subtelomeric methylation changes and telomere length changes during hepatocarcinogenesis. Methylation changes are evaluated as follows: Dmethylation ¼ methylation ratio of HCC À methylation ratio of adjacent non-HCC. Similarly, telomere length changes (Dtelomere) were obtained as follows: telomere length of HCC À telomere length of adjacent non-HCC. Correlation between changes in methylation and those in telomere length was analyzed. p value < 0.05 was marked with *, and a marginal significance (0.1 < p value < 0.05) was marked with **.
CpG methylation and TERRA expression. Detection of TERRA expression in hepatocarcinogenesis and the association of TERRA expression with subtelomeric CpG methylation would provide better understanding of the role of subtelomeric chromatin in hepatocarcinogenesis. Further studies with tumor samples from various tissues (e.g., the breast, colon and gastric tissues) and systematic experimental tests for more subtelomeric regions might clarify the association between subtelomeric methylation and tumorigenesis.
Recently, a negative correlation between subtelomeric methylation frequency and telomere length was reported for several human cancer cell lines. 32 In these studies, the methylation frequency was evaluated by BGS at two subtelomeric sites, and DNA methylation of subtelomeric D4Z4 repeats on chromosome 4q showed a significant negative correlation with telomere length (p < 0.01), while the other site, an independent subtelomeric repeat, SRH, located on chromosomes 1, 9, 15, 16 and X, showed a marginally significant negative correlation (p ¼ 0.07). 32 In our study, correlations between telomere length and methylation ratio were found in certain regions of subtelomeres. Methylation at 7q gradually decreased with telomere lengthening of HCCs, and a similar relationship was found at 18p in both HCCs and non-HCCs, indicating that hypomethylation in these regions might be related to long telomeres. These results are consistent with previous studies, reporting that loss of subtelomeric heterochromatin is concomitant with substantial telomere elongation. 15,33 Interestingly, hypomethylation changes on 7q increased when telomere lengthened from non-HCC to HCC, suggesting that structural change of 7q to relaxed chromatin might be related to telomere lengthening from non-HCC to HCC. Recent studies have indicated that a loss of DNMTs results in increased telomeric recombination, which is a typical feature of ALT cells. 15 It is, thus, interesting to determine whether ALT features such as APBs, ALT-associated PML bodies, increased in telomere elongated HCCs. Conversely, hypomethylation of 21q was frequent in short telomeres of HCCs, and the extent of hypomethylation changes at 21q appeared to be increased when telomere shortened from non-HCCs to HCCs, suggesting that a loss of subtelomeric heterochromatin at 21q might be associated with telomere short-ening. A loss of subtelomeric and telomeric heterochromatic features were found to be related to telomere lengthening in DNMT deficient cells 15 and telomere shortening in telomerase-negative and TRF2 overexpressed cells as well, 16,34 suggesting a diverse role of telomeric chromatin in telomere regulation. It is not clear how subtelomeric methylation in different regions can have positive, negative or no correlation with telomere length in hepatocarcinogenesis. Subtelomeric chromatin in specific regions might differently involve in telomere regulation. For instance, the demethylation of certain subtelomeric regions might induce telomere regulators, e.g., TERRA, that in turn is involved positively or negatively in telomere regulation during hepatocarcinogenesis. We, however, do not exclude a possibility that the average length of the bulk telomere population has misled the results. To elucidate the relationship between telomere length and subtelomeric chromatin, it is needed to determine telomere length at individual chromosomes which can be detected by single telomere length analysis (STELA) 35 and compare the results with subtelomeric CpG methylation determined in extensive regions of the same chromosomes.
It was noted that 21q in HCCs had higher and wider methylation ratio compared to non-HCCs, and 21q showed frequent alteration on the methylation status during hepatocarcinogenesis. Similarly, 7q exhibited frequent methylation changes and a hypervariable methylation ratio in HCCs. Meanwhile, regions on 8q, 18p and XpYp, which showed no great difference in methylation ratio between non-HCCs and HCCs and less-frequent methylation changes, exhibited little correlation with telomere length changes. These results suggest that the subtelomeric regions revealing frequent and dynamic methylation pattern changes might be potential regions whose chromatin structures affect telomere length regulation.
In summary, the subtelomeric methylation patterns dynamically changed during hepatocarcinogenesis. Subtelomeric regions that exhibited diverse methylation ratios in HCCs and frequent methylation change from non-HCCs to HCCs tended to be associated with telomere length regulation in hepatocarcinogenesis. This study demonstrated a potential association between subtelomeric epigenetic modifications and telomere regulation during human hepatocarcinogenesis.